According to the FBI, threat actors have been contacting current and former federal and state officials through fake voice and text messages claiming to be from trusted sources. These scammers then attempt to establish rapport before directing victims to malicious websites to extract passwords and other private information.

Apart from cautioning about the hackers’ tendency to compromise one official’s account, the FBI believes these threat actors may use that access to impersonate the victims further and target others within their network. Verifying identities, avoiding unsolicited links, and enabling multifactor authentication to protect sensitive accounts will be even more crucial.

The FBI and cybersecurity experts are recommending examining media for visual inconsistencies, avoiding software downloads during unverified calls, and never sharing credentials or wallet access unless certain of the source’s legitimacy.

An evolving threat

Essentially, we are talking about scams where sophisticated AI is used to create highly convincing audio, images, text, or videos that look, sound, and act like real people. The easy availability of this technology practically gives fraudsters access to Hollywood-style special effects, enabling bad actors to commit deepfake fraud at scale. The World Bank reports that deepfake fraud has surged by 900% in recent years. Losses fuelled by generative AI are on track to reach $40 billion by 2027.

Deepfake fraud has become troubling because of its highly realistic nature, accessibility to fraudsters, and scalability. Generative artificial intelligence and deepfakes are making existing types of fraud, such as new account fraud, account takeover, phishing, impersonations, and social engineering, even more costly. While voice-cloning deepfakes have successfully targeted several global businesses, video-based deepfakes are empowering criminal groups like the Yahoo Boys with compelling romance scams.

Consider this: Generative AI rapidly creates images that appear ‘realistic’ with almost zero imperfections, eliminating telltale signs of deepfakes such as strange-looking fingers, distorted faces, or stretched-out arms. To make matters worse, using cloud computing, criminals can launch multiple attacks simultaneously or create a large volume of synthetic content for a targeted campaign, such as spear-phishing fraud.

Generative AI and deepfakes are already being incorporated into several common frauds. This includes “New Account Opening Fraud,” where criminals use deepfake technology with synthetic videos, audio, or images that appear to be a legitimate person opening a new bank account. From there, they can bypass facial recognition or liveness detection measures. By mimicking an account holder’s appearance, voice, and mannerisms, fraudsters can convince a customer service representative to grant them access to someone else’s account.

Spelling and grammar mistakes were once obvious red flags of phishing scams. However, thanks to GenAI, criminals are less likely to make these errors. Fraudsters can now craft persuasive phishing messages that are grammatically correct, contextually relevant, and have perfect spelling.

Fraudsters can also convincingly imitate individuals in professional settings, such as meetings or legal proceedings, to commit fraud. In personal settings, they can pretend to be a loved one in need of financial or medical help, as in a romance or grandparent scam. Synthetic identities (fake identities created by combining real and fictitious information) are now appearing to look like real people. These synthetic identities are defrauding businesses and other individuals.

In Hong Kong, a financial worker was tricked into paying out $25 million when fraudsters used deepfake technology to impersonate the company’s CFO. In Italy, a group of entrepreneurs was targeted by scammers earlier in 2025, who copied the Defence Minister Guido Crosetto’s voice and requested money to help pay the ransom of journalists kidnapped overseas.

At least one victim paid €1 million to an overseas account. WPP Digital CEO Mark Read said United Kingdom-based scammers unsuccessfully used a combination of a voice clone and YouTube footage to schedule a meeting with themselves and ad company executives in 2024.

Video-based deepfake frauds make impersonation-based fraud, like romance scams, even more difficult to catch. In 2024, American consumers lost an estimated $1.14 billion to romance scams. With deepfake technology, scammers can create a large library of fake online suitors. Aided by advanced large language models (LLMs) like LoveGPT, romance scammers can target multiple victims at the same time.

Manipulating publicly available images to commit romance scams has proven effective. In 2024, a scammer used simpler technology to deceive a French woman into believing she was in a relationship with Brad Pitt. Organised romance scam groups like the Yahoo Boys are creating more personalised communication for their targets in real time, making romance scams even more convincing and likely to succeed.

Even tech boss Elon Musk couldn’t save himself from being deepfaked. In 2024, there were reports of AI-powered videos posing as genuine footage of the Tesla and X (formerly Twitter) boss going viral. The New York Times dubbed deepfake “Musk, the Internet’s biggest scammer.”

Steve Beauchamp, an 82-year-old retiree, told the New York Times that he drained his retirement fund and invested $690,000 in such a scam over several weeks, convinced that a video he had seen of Musk was real. His money soon vanished without a trace.

“Now, whether it was AI making him say the things that he was saying, I really don’t know. But as far as the picture, if somebody had said, Pick him out of a lineup, that’s him. Looked just like Elon Musk, sounded just like Elon Musk, and I thought it was him,” Beauchamp told the NYT.

Deepfake-powered videos can fuel other impersonation tactics like “CEO fraud” or grandparent scams. If the target believes they are interacting with the real person, they are more inclined to follow their instructions to help their company or a family member.

While audio and visual manipulation have emerged as critical components behind the deepfakes’ success, the rest depends on trust. Here, psychological manipulation from social engineering is working wonders for cybercriminals.

By scouring information like social media profiles, compromised data, or other sensitive information, fraudsters create specific scenarios that emotionally trigger their targets and quickly gain their attention and trust. The more detailed a story the scammer presents, the more believable it is.

Businesses and banks may see a rise in highly personalised “scams as a service” tactics. Criminals can purchase pre-configured deepfake materials for a specific target (a bank manager or executive), in addition to accessing information like email lists to gain intel on any financial organisation’s internal hierarchy.

Money and trust getting eroded

In a 2024 Deloitte poll, 25.9% of executives revealed that their organisations had experienced one or more deepfake incidents targeting financial and accounting data in the 12 months prior, while 50% of all respondents said they expected a rise in attacks over the following 12 months.

The United States Financial Crimes Enforcement Network (FinCEN) issued an alert in 2024 to help financial institutions identify fraud schemes that use deepfake media created with GenAI tools.

The network observed an increase in suspicious activity reports from financial institutions describing the suspected use of deepfake media in fraud schemes targeting their institutions and customers, beginning in 2023 and continuing into 2024.

Deloitte’s Centre for Financial Services predicts that GenAI could enable fraud losses to reach $40 billion in the United States by 2027. To make matters worse, digital trust is “crumbling” under an avalanche of synthetic media, misinformation, and deepfake fraud, according to a new report from Jumio.

The firm’s fourth annual “Jumio Online Identity Study” surveyed 8,001 adult consumers split equally between the United States, Mexico, the United Kingdom, and Singapore. They have much in common: namely, a growing fear that AI-powered fraud now poses a greater threat to personal security than traditional forms of identity theft, and a corresponding rise in skepticism about anything and everything online.

“Fraud-as-a-service (FaaS) ecosystems have erupted like a bad rash, enabling even amateur fraudsters to leverage synthetic identities, deepfake videos, and botnet-driven account takeovers. Consumers must navigate scam emails, manipulated social media content, and digitally altered identity documents. Seven out of ten global consumers (69%) indicated they are more skeptical of the content they see online due to AI-generated fraud than they were last year,” the report noted.

When asked who they trust most to protect their personal data, 93% of respondents said they trust themselves over the government or Big Tech.

However, Jumio said, “Self-reliance does not mean consumers want to go it alone. In fact, when asked who should be most responsible for stopping AI-powered fraud, 43% pointed to Big Tech, compared to just 18% who chose themselves.”

The research further showed that consumers are open to modernised fraud protection, even if it means additional steps. Most respondents globally said they would be willing to spend more time completing comprehensive identity verification processes, especially in sectors where the stakes are high, like banking or healthcare.”

But it also recognises that technology alone is not the answer. Jumio CEO Robert Prigge said, “Building a trustworthy digital world depends on strong consumer education and transparency. With day-to-day worries about generative algorithmic technologies on the rise, the trust gap also continues to grow proportionally. As such, businesses must also earn consumer trust in these protections.”

The age of paranoia kicks in

Nicole Yelland, who works in public relations for a Detroit-based nonprofit, now conducts a multi-step background check whenever she receives a meeting request from someone she doesn’t know. Yelland runs the person’s information through Spokeo, a personal data aggregator. If the contact claims to speak Spanish, Yelland says, she will casually test their ability to understand and translate trickier phrases. If something doesn’t quite seem right, she’ll ask the person to join a Microsoft Teams call— with their camera on.

If Yelland sounds paranoid, that’s because she is. In January, before she started her current nonprofit role, Yelland says, she got roped into an elaborate scam targeting job seekers. “Now, I do the whole verification rigmarole any time someone reaches out to me,” she said to WIRED.

In a time when remote work and distributed teams have become commonplace, professional communication channels are no longer safe, thanks to the GenAI-powered scams. The same AI tools that tech companies use to boost worker productivity are also making it easier for criminals and fraudsters to construct fake personas in seconds.

Big Tech journalist Lauren Goode said, “On LinkedIn, it can be hard to distinguish a slightly touched-up headshot of a real person from a too-polished, AI-generated facsimile. Deepfake videos are getting so good that longtime email scammers are pivoting to impersonating people on live video calls. According to the US Federal Trade Commission, reports of job and employment-related scams nearly tripled from 2020 to 2024, and actual losses from those scams have increased from $90 million to $500 million.”

Yelland says the scammers who approached her in January 2025 were impersonating a real company, one with a legitimate product. The “hiring manager” she corresponded with over email also seemed legit, even sharing a slide deck outlining the responsibilities of the role they were advertising.

However, during the first video interview, Yelland says, the scammers refused to turn their cameras on during a Microsoft Teams meeting and made unusual requests for detailed personal information, including her driver’s license number. Realising she’d been duped, Yelland slammed her laptop shut.

These schemes have forced AI players to work on technologies to detect other AI-enabled deepfakes, including GetReal Labs and Reality Defender. OpenAI CEO Sam Altman also runs an identity-verification startup called “Tools for Humanity,” which makes eye-scanning devices that capture a person’s biometric data, create a unique identifier for their identity, and store that information on the blockchain. The whole idea behind it is proving “personhood,” or that someone is a real human.

“A section of corporate professionals is also turning to old-fashioned social engineering techniques to verify every fishy-seeming interaction they have. Welcome to the age of paranoia, when someone might ask you to send them an email while you’re mid-conversation on the phone, slide into your Instagram DMs to ensure the LinkedIn message you sent was really from you, or request you text a selfie with a time stamp, proving you are who you claim to be. Some colleagues say they even share code words with each other, so they have a way to ensure they’re not being misled if an encounter feels off,” Goode stated.

Daniel Goldman, a blockchain software engineer and former startup founder, said, “What’s funny is, the lo-fi approach works.”

Goldman began changing his own professional behaviour after he heard a prominent figure in the crypto world had been convincingly deepfaked on a video call.

He ended up warning his close ones that even if they hear “his voice” or “see him” on a video call asking for money or an internet password, they should hang up and email him first before doing anything.

Ken Schumacher, founder of the recruitment verification service Ropes, has worked with hiring managers who ask job candidates rapid-fire questions about the city where they claim to live on their resume, such as their favourite coffee shops and places to hang out. Another verification tactic being used by people is what Schumacher calls the “phone camera trick.”

Here, if someone suspects the person they’re talking to over video chat is being deceitful, they can ask them to hold up their phone camera to show their laptop. The idea is to verify whether the individual may be running deepfake technology on their computer, obscuring their true identity or surroundings.

However, it’s safe to say this approach can also be off-putting: Honest job candidates may be hesitant to show off the inside of their homes or offices, or worry a hiring manager is trying to learn details about their personal lives.

“Everyone is on edge and wary of each other now,” Schumacher says, and it perfectly sums up the mood change people are undergoing in the age of GenAI-powered scams.

As deepfakes grow more advanced and accessible, AI-driven scams are reshaping cybercrime. Traditional security is no longer enough; vigilance, identity checks, and robust cybersecurity frameworks are the need of the hour to counter this rising threat.

The post Deepfake fallout: Welcome to the age of paranoia appeared first on International Finance.

Often, without continuous human oversight, these artificial intelligence bots may spot issues, create and weigh several answers, decide what to do, and carry out judgments. Particularly for its ability to transform how operators maximise infrastructure, solve network anomalies, lower operational loads, and improve customer experience with intelligent self-service technologies, this paradigm shift is generating great interest in the telecom industry.

Notwithstanding the futuristic promise, a fundamental concern remains: Is the telecom sector, which has always been hesitant to absorb unproven innovation and highly wary about losing control, ready to embrace a kind of artificial intelligence that questions its conservative DNA?

A synopsis of history

Plotting the history of artificial intelligence in telecoms helps one to see the relevance of agentic artificial intelligence. Early in the 2000s, phone companies started using rule-based systems for fraud detection and automated call routing.

Predictive analytics and machine learning first began to be applied to project network congestion, client turnover, and service demand by the 2010s. More recently, GenAI solutions have helped with content generation and natural language processing for consumer-facing applications.

However, their lack of contextual awareness and real-time execution capability hampered these instruments. Agentic artificial intelligence marks a new horizon whereby autonomous systems may not only evaluate but also act with agency, judgment, and adaptation.

Juniper Research’s observations suggest that agentic artificial intelligence has great potential in telecoms since it may assume challenging roles with less control. Network optimisation is one of the most important uses because agentic systems surpass mere problem prediction.

Through constant learning from patterns and adaptation, these agents may proactively fix network imbalances and preserve optimal performance conditions. Both security and fraud detection benefit equally; rather than depending on fixed rules, these artificial intelligence agents dynamically change security protocols to counter changing cyber threats, thus enhancing resilience in an era of ever-advanced attacks.

Virtual assistants driven by restricted GenAI or chatbot logic usually cannot independently address problems in customer service. Agentic AI, on the other hand, is poised to transform this field by allowing bots to not only comprehend but also act on sophisticated consumer needs, including account settings, refund requests, or technical problem solutions without agent escalation. Long praised but rarely smooth, personalisation becomes more flexible and responsive as AI systems continuously learn and customise services to real-time user behaviour.

Agentic artificial intelligence has a transforming edge in its change from static automation to autonomous orchestration. This new degree of capability brings contextual awareness and decision agency, therefore transcending artificial intelligence’s usual job as supporting analysis in the field of digital operations management. One European CTO said at MWC 2025, “Agentic AI is the brain we always wanted for our networks, but we’re still not sure if we trust it with the steering wheel.”

Existing telecom AI vs agentic AI

Although the phrase “agentic AI” sounds like a futuristic abstraction, its basic ideas are not wholly novel. For years, telcos have been testing network automation, particularly via Self-Organising Networks (SON) in the Radio Access Network (RAN), where artificial intelligence controls network tuning and resource allocation. Still, SON systems are essentially rule-based and reactive. Rather than making decisions on their own, they respond to premeditated stimuli.

Agentic artificial intelligence marks a further development. Targeting predictive correction and autonomous learning, these agents are not only reacting to events but also constantly scanning, hypothesising, and testing network configurations in the background.

This can cover application performance management, real-time orchestration across RAN and core networks, and anomaly prevention before user impact. The intricacy and seeming risk involved in handing important operational choices to nonhuman systems make the telecom sector understandably dubious, even with great theoretical appeal.

Industries like banking have been more aggressive with artificial intelligence autonomy by comparison. Though dangers vary, algorithmic trading in finance or smart manufacturing with dynamic supply chain changes shows that agentic systems are possible. The difference emphasises how adoption curves are shaped by sector-specific pressures and legal obligations.

ROI expectations and economic impact

Adopting agentic artificial intelligence has significant financial consequences, rather than only technical ones. Over the next ten years, industry observers estimate that intelligent automation in telecoms might release up to $200 billion in value. Improved client retention, better network use, and operational cost savings could account for a good portion of this value.

While tailored customer journeys driven by AI may boost ARPU (average revenue per user), AI-based predictive maintenance could help lower downtime and related service fines. Reducing the requirement for customer service escalation or manual network troubleshooting could also free up important human resources for more high-value projects. These efficiency improvements could be very important for long-term viability for carriers running slim margins.

Transparency in making ethical decisions

Ethical issues will surely become more pressing as agentic artificial intelligence finds an increasing presence in telecom operations. Should AI agents be allowed to provide traffic top priority during network congestion, and if so, under what standards? Might personalising algorithms that favour high-revenue consumers disadvantage some user groups? When AI decides on choices with social or political ramifications, including the undervaluation of connectivity in underdeveloped areas, what follows?

To meet these issues, openness and explainability will be crucial. Telcos will have to create systems that clearly justify their autonomous judgments as well as make them. The system design philosophy has to change from blackbox models toward interpretable and auditable AI systems. As artificial intelligence’s influence in telecom services grows, regulatory authorities will probably seek more responsibility.

Learnings from other domains

Telecoms have been quite cautious when implementing autonomous artificial intelligence systems compared to other industries. Algorithmic trading systems now make choices in milliseconds and run with almost no human involvement in finance. Smart manufacturers utilise artificial intelligence to dynamically change supply chains and maximise manufacturing processes depending on demand and logistics.

Often in high-stakes settings, even healthcare is starting to adopt AI-powered diagnostics and patient monitoring.

These industries present both successful blueprints and cautionary stories. Successful adoption of artificial intelligence depends on progressive implementation, well-defined return on investment, and a significant emphasis on ethical monitoring, among other common denominators.

Telecoms have to absorb these lessons by building trust through openness, implementing human-in-the-loop protections, and matching technology change with cultural and legal realities if they are to properly leverage agentic artificial intelligence.

Future applications beyond the standard

Although present agentic artificial intelligence applications in telecoms mostly address RAN optimisation, network performance, and customer service, their possibilities go much further. Agentic artificial intelligence could spread into more significant corporate roles in the next few years.

Dynamic pricing engines, for example, might independently change service plan rates using real-time network statistics and market trends. AI agents may also assist operators in managing spectrum allocation depending on changing traffic and geopolitical conditions or in reacting to the marketing of competitors.

Agentic artificial intelligence might also be involved in marketing, doing A/B tests on offers, and independently scaling the most successful campaigns or changing messaging depending on live sentiment analysis.

Autonomous systems could find abnormalities in bills and send proactive communications or corrections to consumers without human interaction. The potential of AI models to impact high-level business results will likely increase in parallel as they develop and train on multimodal data (from speech, usage behaviour, location, and CRM inputs).

Industry viewpoints are still split, but they are also growing interesting. Some CEOs on the ground at events like MWC 2025 predict agentic AI as unavoidable. Others exercise caution, however.

A Chief Technology Officer from a Scandinavian operator remarked, “We’re not looking to replace our engineers, but why not let an artificial intelligence spot and fix abnormalities faster than a NOC team?”

Analysts also agree: telcos must lean in rather than jump. Emphasising “measured experimentation” as a strategic approach to agentic AI, Gartner’s 2025 telecom tech view exhorts operators to “fail fast in contained zones” to build trust without losing stability.

Right now, the largest divide resides not in technology but in trust. Cultural changes and organisational buy-in will be essential to scale the promise of artificial intelligence from hype to daily use as agentic systems gradually establish themselves in constrained roles.

Execution is elusive

Although concrete, large-scale projects are still a year or two off, with 2026 expected as a likely turnaround point, some telcos have already dipped their fingers in the water. For example, recently, Google Cloud and Deutsche Telekom revealed a cooperation to test agentic artificial intelligence in a RAN context.

Without waiting for human input, this artificial intelligence agent is designed to spot performance problems in real-time and act immediately for correction. Enhanced network uptime, lower costs, more bandwidth stability, and better client experiences are among the expected results.

In the same vein, Telenor and Ericsson co-developed a proof-of-concept aimed at maximising RAN capacity and energy economy. Particularly in changing demand environments, a rising issue in 5G and soon-to-be 6G infrastructures, their usage of agentic AI promises to balance traffic loads more efficiently.

Industry insiders warn that these initiatives are still in controlled conditions and distant from being implemented over commercial networks, despite the hoopla around events like MWC 2025.

The unifying thread running throughout these early-stage initiatives is that agentic artificial intelligence is intriguing but still immature. Important unresolved issues still include performance under erratic pressure, interface with legacy systems, and openness of autonomous judgments. AI without explainability in decision paths could cause compliance problems.

Operators still demand authority

According to a recent Telecoms.com survey on 5G AI adoption, reliability, responsibility, and service assurance still define the telecoms sector. Only 10% of Communication Service Providers (CSPs) today feel comfortable introducing unsupervised, totally autonomous AI into live production environments.

About 78% of respondents would rather have partially automated or supervised solutions, keeping human operators in the loop.

Why the opposition? Three main obstacles emerged: a lack of in-house AI expertise to oversee these complex systems, the financial cost of upgrading infrastructure to fit advanced AI workloads, and most importantly, a strong worry about dependability and trust.

Telecom infrastructure is too important for systems to fail or make opaque decisions that humans cannot readily audit or undo; it is even sometimes life-saving. This emphasises a major drawback in the current drive for agentic artificial intelligence: operator confidence is not where the technology may be ready.

Ethical questions add still another level of complication. Should artificial intelligence be allowed to give some clients higher priority than others under bandwidth constraints? How can personalisation driven by artificial intelligence prevent aggravating digital inequality? The more decisions these systems make, the greater the demand for accountability mechanisms to monitor and explain them.

Will the leap be taken?

From AI-assisted systems to agentic designs, this is an ideological as much as a technical challenge. It challenges accepted wisdom regarding responsibility, control, and the part human knowledge can play in intricate systems. This leap forces carriers to fundamentally rethink operations, far more than a small step.

And there are significant hazards as well. Autonomous agents that misinterpret scenarios could cause operational interruptions, expose security flaws, or conflict with laws requiring human supervision. Add to this the increased scrutiny from authorities and watchdogs all throughout Europe, North America, and Asia, and it is evident why telcos are still reluctant to fast-track this change.

Still, small adoption could be the path ahead. Telcos could begin with hybrid systems, semi-agentic configurations whereby AI agents propose and carry out low-risk decisions under supervised observation. Before extending to mission-critical areas like core infrastructure or emergency communication layers, this hybrid approach provides a safe runway for internal familiarity and trust-building.

Over the next decade, industry analysts project that intelligent automation in telecoms could generate as much as $200 billion in value. Although only a fraction of this can be attributed to agentic artificial intelligence, its influence is projected to grow as systems develop and confidence increases.

Evolution, not revolution

Agentic artificial intelligence seems on paper to transform telecom operations. Real-world acceptance is probably not going to be sudden, though. Rather, the change will probably be seen as a slow evolution whereby companies add agentic elements on top of current systems instead of replacing them entirely.

Still, a change is in process. Over half (51%) of CSPs in another Telecoms.com survey said that artificial intelligence is starting to profoundly change their perspective on operations, customer interaction, and service delivery.

These figures indicate growing excitement mixed with institutional and cultural immobility. Particularly in layers of artificial intelligence operations and customer management, where agentic capabilities can generate quantifiable returns without compromising fundamental service dependability, investments are gradually rising.

Concurrently, telecom decisions on new technologies are being shaped by outside elements ranging from geopolitical concerns to inflation-driven cost management measures. Even if customer-facing installations remain delayed, the urge to safeguard systems through agentic techniques may ironically hasten adoption in some security roles in areas with weak data laws or dangerous cyber environments.

Agentic AI may find its way from network administration into strategic tasks, including infrastructure investment forecasts, market competitiveness response, or real-time price optimisation in the long run. Still, those use cases are only conceptual. For telecom companies, agentic artificial intelligence has great promise, offering a means to turn reactive systems into intelligent, proactive networks. But telecoms is a business based on dependability, consistency, and confidence.

Agentic systems may shine in demos and pilot environments, but to have a significant role in telco operations, it will demand a demonstrated track record of dependability and openness.

In the near term, one might expect a pragmatic approach: well-scoped pilot projects, semi-autonomous decision-making in controlled situations, and a consistent change in perspective from human-led to human-supervised AI.

The actual change will be incremental and consistent, not a flash revolution. Agentic artificial intelligence is still a fascinating frontier for now, and one operator will explore with one hand on the throttle and the other firmly on the brake.

The post The $200 billion potential of agentic AI in telecoms appeared first on International Finance.

The basic argument of this discussion is a strong one: artificial intelligence will someday replace 90–95% of the coding tasks now performed by humans. Does this indicate, therefore, that we are approaching a day when software programmers will be extinct? Alternatively, are we just reinventing what their duties in the era of artificial intelligence look like?

Is AI The New Intern?

Recently, Microsoft CTO Kevin Scott said, “95% of coding will be AI-generated”. Not only controversial, Scott’s forecast is based on the obvious advancement of AI systems including Meta’s Code Llama, Google’s AlphaCode, and GitHub Copilot (run on OpenAI’s Codex).

These algorithms can suggest fixes, write syntactically perfect code, and even debug current codebases. For everyday coding chores such as boilerplate code, CRUD operations, and simple scripts, these technologies currently exceed junior software developers in both speed and cost-efficiency.

Similar ideas are expressed by Sridhar Vembu of Zoho. He anticipates a tech sector in which demand for conventional software engineers, particularly for entry-level jobs, will sharply drop. His case is this: why pay legions of engineers to accomplish what might be done if AI can write most of the codes? Companies may instead concentrate on selecting few but more strategic engineers with knowledge of efficient AI tool orchestration.

Still, this vision is not perfect. GenAI models like AlphaCode and Codex are not perfect even if they can generate code really remarkably.

AI models battle with contextual awareness, first of all. Most practical software issues are deeply ingrained in business logic, changing requirements, and legacy architectures; they are not isolated puzzles. At most, GenAI understands these subtleties quite superficially.

Second, typically the most difficult aspects of software development—debugging and refactoring—are where artificial intelligence fails. Models can hallucinate and provide syntactically accurate but logically incorrect outputs or outputs incompatible with the constraints of the situation.

Third, security and performance optimisations still require a human eye. Unless specifically taught on such datasets, AI does not naturally grasp security flaws. Even then, it may overlook zero-day issues or unique attack paths.

Context, Creativity, And Critical Thinking

The trio of human strengths—context, creativity, and critical thinking—defines the core of the case against artificial intelligence substituting for software programmers.

Though it cannot yet construct sophisticated systems with an understanding of long-term maintainability or inter-team coordination, AI can create code. It cannot appeal why a product should operate one way only depending on user psychology or market conditions instead of another. Furthermore, in high-stakes industries like defence, fintech, and healthcare, the margin for error is almost non-existent. Here, human supervision is absolutely necessary rather than discretionary.

AI models are also only as good as the data they are trained on. They carry the prejudices, restrictions, and antiquated methods found in their training materials. Software developers run the danger of spreading hidden vulnerabilities and technical debt if they mindlessly embrace AI-generated answers.

Many experts argue for a hybrid future whereby software engineers collaborate with AI rather than seeing it as a substitute.

See AI as a power tool. GenAI can maximise engineers, much as a circular saw made carpenters more efficient rather than eliminated them. AI can be assigned tasks such as rapid prototyping, documentation generation, and unit testing, thereby freeing engineers to concentrate on architecture, performance, and creativity.

When Kevin Scott discusses artificial intelligence as a productivity booster, he makes hints toward this. Sam Altman has also underlined the value of human-in-the-loop systems, whereby AI functions as a co-pilot rather than a captain.

Implications follow from this. Tomorrow’s software engineers must become experts in AI-assisted processes. Understanding prompt engineering, model constraints, and ethical issues will be more important than knowing how to write neat code.

What Businesses Could Do?

From a commercial standpoint, the impulse to automate development in order to save engineering expenses is reasonable but maybe ill-founded.

For basic applications, depending just on artificial intelligence for software development could be a short-term solution. But when systems get more sophisticated, the shortcomings of present GenAI models become clearer. More importantly, AI-generated code still needs review, integration, testing, and deployment—activities best left to professional programmers.

Hiring engineers that are AI-literate is the more environmentally friendly approach. These experts can help artificial intelligence generate excellent code faster and guarantee its fit with corporate objectives. Businesses which make investments in GenAI technologies to upskill their staff will probably have a competitive advantage in the not-too-distant future.

Unquestionably, artificial intelligence is changing the scene in software engineering. Still, it is changing the nature of their work instead of making engineers obsolete.

The winners in this new paradigm will not be those who oppose AI nor those who mindlessly support it. Rather, the benefit will go to software engineers and companies who discover the sweet spot, where human and artificial intelligence will mix together to produce code that is not just efficient but also ethical, future-proof, and robust.

Hence, no, artificial intelligence cannot really replace programmers. At least not yet. Maybe never quite in the way intended here. Still, it can replace those who refuse to grow. We are living in the time of the AI-augmented developer. Will you adapt or be replaced is the question.

The post IF Insights: Is AI a real threat to software developers? appeared first on International Finance.